AI statement and crawler compliance
Benchmarq is committed to transparent, deterministic, and responsible diagnostic engineering. Our audit core does not use generative AI to guess audit scores. We evaluate websites against 200+ rule-based checks derived from verified crawler protocols, structured data standards, and European AI governance frameworks.
Synopsis
- Audit methodology
- 100% deterministic rule assertions. Zero generative AI hallucinations in test evaluations or benchmark calculations.
- Third-party AI sharing
- Zero customer data or scanned content is transferred to third-party AI models for training or distillation.
- Automated decisions
- None. Audit reports provide technical guidance to empower human engineering and SEO teams.
- Human oversight
- All 200+ rules, severity weights, and diagnostic criteria are engineered, reviewed, and audited by human architects.
- EU AI Act
- Our scoring uses deterministic, human-written rules, not AI models. If we add AI-based features, we will disclose them as the EU AI Act requires.
Deterministic rule assertions vs generative guessing
Why Benchmarq uses programmatic validation instead of stochastic Large Language Models to calculate readiness.
Reproducible benchmark scores
Generative LLM prompts produce inconsistent outputs across repeated runs. Benchmarq's Discovery & Authority Engine (DAE) executes deterministic rule assertions: identical code yields identical scores across every scan.
Read scoring methodology →Inspectable code-level causes
Every point deduction links directly to a failed check (e.g. missing JSON-LD schema, blocked robots.txt directive, or malformed cache headers) with concrete code diffs.
Inspect audit engine →Isolated specialist modules
Advanced add-ons like Agentic Commerce and Agentic Apps run as isolated audit suites, ensuring specialized protocol tests never distort your baseline AI readiness benchmark.
Explore ecommerce suite →Data isolation and model protection
Strict guarantees ensuring your proprietary source code and scan telemetry remain completely isolated.
Zero training on client data
We never feed scanned domain URLs, page structures, benchmark scores, or private action items into public or private LLMs for model training.
No third-party LLM queries
Benchmarq does not send your data to external generative API providers. If specialized LLM evaluations are introduced in the future, they will be explicitly disclosed with granular opt-in controls.
No personal data harvesting
The scan reads only publicly rendered markup, schemas, and headers. That material is what the AI readiness audit scores.
Human oversight and rule governance
Every check in the Benchmarq registry is created and maintained by experienced software and search engineers.
Expert rule authoring
Diagnostic tests are built by search engineers and digital architects with 20+ years of experience in semantic indexing, web performance, and API design.
Continuous standards tracking
We monitor official crawler releases (GPTBot, ClaudeBot, PerplexityBot, Google-Extended) and the specifications our checks rely on (W3C standards, Schema.org, and protocols such as Google's AP2) to update test criteria as search architectures evolve.
Review release notes →Ethical crawler etiquette
Our audit worker respects HTTP rate limits, robots.txt directives, and server load thresholds when crawling entire sitemaps.
Questions regarding our AI policies?
Contact our technical team for architecture details.